Identification and Characterisation of Infiltrating Immune Cells in Malignant Pleural Mesothelioma Using Spatial Transcriptomics

Abstract

Increasing evidence strongly supports the key role of the tumour microenvironment in response to systemic therapy, particularly immune checkpoint inhibitors (ICIs). The tumour microenvironment is a complex tapestry of immune cells, some of which can suppress T-cell immunity to negatively impact ICI therapy. The immune component of the tumour microenvironment, although poorly understood, has the potential to reveal novel insights that can impact the efficacy and safety of ICI therapy. Successful identification and validation of these factors using cutting-edge spatial and single-cell technologies may enable the development of broad acting adjunct therapies as well as personalised cancer immunotherapies in the near future. In this paper we describe a protocol built upon Visium (10x Genomics) spatial transcriptomics to map and characterise the tumour-infiltrating immune microenvironment in malignant pleural mesothelioma. Using ImSig tumour-specific immune cell gene signatures and BayesSpace Bayesian statistical methodology, we were able to significantly improve immune cell identification and spatial resolution, respectively, improving our ability to analyse immune cell interactions within the tumour microenvironment.

Keywords: immune checkpoint inhibitors; mesothelioma; research protocol; spatial analysis.

Conflict of interest statement

The authors declare no conflict of interest.

Figures

Figure 1

Overview of Visium spatial analysis workflow. Sections from FFPE tissues chosen for the spatial transcriptomic workflow are first quality checked, then freshly cut sections are adhered to the Visium slide. Following deparaffinisation and staining, they are imaged using a slide scanner followed by library construction, PCR amplification and sequencing. Spatial data are finally analysed using the Loupe Browser or other bioinformatic techniques as we describe herein. H&E—Hematoxylin & Eosin; UMI—unique molecular identifier.

Figure 2

H&E and spatial images of analysed tissues using Loupe browser. H&E images of tissues A1 and B1 scanned with Olympus slide scanner (A,E) and pathologist’s annotation of tissues A1 and B1 in Loupe browser (B,F). To identify immune cells within the tissue sections, pan-T-cell (C,G) and pan-B-cell (D,H) markers were queried in Loupe browser. Discordance of spots assigned to T and B cells using pan-cell markers represent a major limitation of cells identification in Loupe browser.

Figure 3

Identification of tumour-infiltrating immune cells using the ImSig algorithm in tissues A1 and B1. (A,D,G,H) Loupe images demonstrating infiltrating immune cells identified using the ImSig algorithm in tissue A1 and B1 respectively. (B,E) R images representing the selection of custom cut-off values for each type of immune cell(s). (C,F) Histograms represent the cut-off values assigned to each subtype of immune cells using Seurat function. AddModuleScore was used to score features for signatures and made a manual cut-off for each signature to exclude the negative population (a la flow). (I) Heatmaps representing the top 50 significantly upregulated genes in T-cell spots compared to non-T-cell spots from sections A1 and B1. The relative expression of these genes in other Imsig-based immune cell spots indicates that there is significant overlap in B-cell and T-cell identification due to non-single cell resolution of the Visium slide. (J) Pathways enriched by the top 50 differentially expressed genes in T cells of tissues A1 and B1. Again, we can observe an overlap with B-cell related pathways emphasising the limitation of the Visium spatial method.

Figure 4

Increased resolution of spatial images with BayesSpace method. (A,B) Enhanced resolution allowed us to more precisely locate and quantify tumour-infiltrating immune cells, particularly T cells, B cells, NK cells and macrophages using the ImSig algorithm. (C,D) Scales and histograms represent the cut-off values assigned to elucidate the signature scores for each section and/or cell type using an outlier analysis (hampel filter) [20].

Figure 5

Analysis of BayesSpace-enhanced immune cell composition. (A) Imsig-characterised immune cell composition of BayesSpace enhanced tissue sections A1-D1 showing differences between tumour and adjacent non-tumour tissue. (B) Proportion of ImSig-characterised immune cell spots sharing proliferation and interferon signatures in tumour and adjacent non-tumour tissue. (C) Analysis of T- and NK-cell neighbouring spots among different immune cell populations in tumour and adjacent non-tumour tissue.